This paper is focused on real time detection of satellite reception state. In constrained environment, such as urban areas, GNSS signals can be received directly, reflected or blocked by obstacles (building, vegetation, etc) and can lead to an error or a lack of positioning. This paper proposes to characterize the GNSS signals reception environment using image analysis. The proposed approach consists of detecting the visible sky part in the images thanks to a k-means based classification algorithm applied in segmented/simplified images. In this paper, two images segmentation/simplification techniques are proposed and compared in terms of robustness, reliability and processing rate. The first one is based on a combination of color and texture information. The second technique uses a geodesic reconstruction by dilatation with an optimal constrast parameter. The main aim of this work consists of determining how many satellites are positioning in the area of visible sky and have a direct reception state. Experimental classification and positioning results, using real data and an evaluation methodology are presented to demonstrate the effectiveness and the reliability of the proposed approach.